Beethoven’s ninth; an iconic symphony. What are the differences between different versions of this symphony?

Beethoven’s ninth, a piece commonly known, can have a lot of different meanings for different persons. Some people only find the music soothing and beautiful, while other people use it as an identifier, since it’s the anthem of the European Union. Considering this year has brought us Brexit and realising that Beethoven was born 250 years ago this year, gave me a good reason to investigate this masterpiece of a symphony. On Spotify there are countless different versions of this piece. All performed by different conductors with different orchestras. I wanna know what the differences are between these versions. Performing a classical piece like this one doesn’t come with a lot of freedom in interpretation. I wonder whether Spotify statistics are able to identify the small differences between versions. To investigate this i implemented the data of 36 different versions in R. The features i used are danceability, energy, loudness and tempo. Danceability might seem strange in the context of a symphony, however in my research i use it not to measure how ‘danceable’ a track actually is, but to add a variable to my comparison to get a better view of the symphony as a whole. I think the way Spotify constructs the danceability feature and especialy the different elements they look at are interesting enough to use in my research. The next feature is energy. With this feature i try to measure the activity and intensity between the different versions. I’m also gonna measure the loudness of the different versions. Loudness is less of an identifier in my research but it could make differences or similarities between the different versions more conclusive. The last feature i used is tempo. This feature analyses the song by bpm. Spotify takes the mean of the tempo throughout the song. This could result in slightly incorrect data, since the symphony contains a lot of tempo changes. Nevertheless i think it can be very useful in drawing a comparison between the different versions.

Ludwig van Beethoven, by Joseph Karl Stieler, 1820

How does one analyse all these features of the same symphony and compare them to each other?

To better analyse the different versions of the symphonies i took the means and standard deviations of the features of the symphonies. Below you see a table with all those variables.

This table is of course a very raw one and it’s hard to draw any conclusions from it. We can play around with the ordering function a bit, to see which of the versions excels at certain features, but it’s not possible to spot any correlation between different versions.

How to prove correlations: The truth is in the graph

Here is an example of a bar graph considering the difference in tempo between the different versions. Also the standard deviations of the tempo of the different versions is implemented in the graph. This graph shows a lot more than the table that we saw before. There is a clear variety between the different versions. Also there are some outliers like the version by Hans Schmidt Isserstedt, The London Symphony Orchestra with Bernard Haitink and the one by Tafelmusik Baroque Orchestra.

How do the outliers from the previous slides compare, considering timbre?

Shown here are the two cepstograms from the outliers. From this graph there is not much difference to notice. An explanation could be that the orchestra’s were very like-minded considering timbre. Another explanation could be that there are certain standards attached to recording an orchestra and that the differences in interpretation are not hearable.

Key and Chord estimations: Do all orchestra’s follow Beethoven’s sheet music?

These two graphs compare the similarities in key between the two outliers, the version by Tafelmusik Baroque Orchestra and by Hans Schmidt Isserstedt. As is to be expected the graphs are very similar. However the version ### Practice with cluster analysis